Summary: ROSS Hypersparc CPU

From: Robert H. Moyer (rhmoyer@mmm.com)
Date: Sat Feb 04 1995 - 04:28:23 CST


A couple of weeks ago I asked:

> One of the groups that I support has purchased a set of quad Ross Hypersparc
> 90 mhz cpu modules for a 670 server. Before I install these I wonder if anyone
> has had any experience with these that they would like to share, specifically
> any "gotchas" or things that they forgot to tell you in the documentation.

and recieved responses from the following:

Steve Young (syoung@cs.buffalo.edu)
Yves Morin (Yves_Morin@bcomeau.hydro.qc.ca)
Gary L. Berger (Gary.Berger@eng.clemson.edu)
Chris Terry/Computervision
Daniel Strick (dan@bellcore.com)
dave (dave@solar.exar.COM)
Ross Bennett (ross@gcs.oz.au)
rnf@spitfire.tbird.com (rnf@spitfire.tbird.com)
leg@inel.gov (leg@inel.gov)
David B. Brown (dbrown@lizzard.med.utah.edu)
Sanford Whitehouse (sanford@lsil.com)

Most of the answers mentioned the need to install the Ross supplied patches for
4.1.3 or else install 4.1.4 or 2.3 which include native support for the
modules. One person mentioned that they ran hot and a lot of cooling was needed
and a couple mentioned problems that they had problems but they were generally
minor or easily resolved.

With that information in hand I proceeded forward with the upgrade, since I was
not able to get a copy of 4.1.4 I decided to upgrade to 4.1.3_u1 and then
install the Ross patches. Anyway, this turned into a long weekend, here is the
message that I sent to the list a couple of days later.

---------
Over this past weekend I upgraded our Sun 670, the operating system was upgraded
from 4.1.3 to 4.1.3_U1 and I replaced the single Sparc processor with four Ross
Hypersparc processors. The results were relatively positive with the exception
of two nagging problems that I have been upable to resolve, both of which seem
to relate to loss of support in the kernel for some unbundled software.

1) After upgrading to 4.1.3_U1 I lost support for DiskSuite 1.0. Fortunately,
   we were not using it for striping or concatenation, just mirroring and
   so the system is operational I just have no mirroring. The 4.1.3_U1 doc
   lists this as supported software and a search of the Sunsolve databases,
   FAQ's, etc. did not reveal any particular problem reports related to this. I
   have reinstalled DiskSuite but still no luck. I am using the same kernel
   configuration file that I did before the upgrade and it still contains
   the line "pseudo-device md128" for DiskSuite support and the 3 lines
   in conf/files still exist.

   The error I recieve bootup is "Unable to open /dev/rmdd, No such
   Device or Directory". Yes, the device exists, the DiskSuite manual
   tells me that this message means that DiskSuite is not installed properly.
   Has anyone else experienced this problem before and been able to successfully resolve it?

2) After patching the new 4.1.3_U1 kernel with the Ross patches and installing
   the modules I lost support for my ALM-2 boards. Fortunately they are not
   providing a critical service for the system's operation but they are used
   heavily off-hours for dial-in support. Again the kernel config file has
   not changed, any ideas out there?

--------------

I did not recieve any responses to this request or at least I do not think
I did because a couple of days later I realized the sendmail was not working
properly and it is very possible that reponses were either bounced back to
the sender or sent to the bit bucket.

Anyway, the second problem was resolved after a call to ROSS revealed that the
Dataram memory expansion board that the client had installed was incompatible
with the Ross Modules. Removing the simms from the Dataram board and moving
them to the system board cleared up the problem and gave us access to the
ALM boards again. Interestingly enough ROSS told us that we were the only
people to have ever gotten the system to boot at all with the combination
of boards installed.

The first problem I sort of resolved when I found that conf.c did not include
the lines to support DiskSuite even though I had reinstalled the software. It
seems that the 4.1.3_u1 stepped on this file. I manually inserted the
lines to get the necessary support in the kernel but just rebooting with the
new kernel and reinitializing the metadevices with metainit caused severe
problems on the system that required a number of the filesystems to have to be
completely restored. It turns out here that the correct steps for me to take
would have been to rebuild all the metadevices from scratch as if they had
never existed on the system at all which I am now in the process of doing.

Both of the problems that I outlined here were known by the vendor
(ROSS and Sun) but neither were documented in their manuals or release
notes. So thanks to all the folks on this list who either responded or tried
to respond to my requests but no thanks to the vendors who failed to supply
me with the information that I needed to make this an easy upgrade.

-- 
                                                         _/_/_/ _/    _/
Robert Moyer                                                _/ _/_/_/_/
3M - IT/EIS Systems Support                              _/_/ _/ _/ _/
rhmoyer@mmm.com                                           _/ _/    _/
                                                     _/_/_/ _/    _/ Innovation.



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:15 CDT